INTRODUCTION TO (OPERATOR ALGEBRA) K-THEORY, JAMIE WALTON, OPERATOR ALGEBRAS WORKING SEMINAR, 26TH OCTOBER 2018 1. Origins K-theory for topological spaces was developed by Atiyah and Hirzebruch for topological spaces in 1959, with roots in earlier work by Grothendieck in algebraic geometry. K-theory for C∗- algebras began to be developed in the 1970s. Although our interest is in C∗-algebras, let's very briefly see how the construction goes for topological spaces (there are important relations between the two approaches). Let X be a compact, Hausdorff space; a minor modification is needed if the space is only LCH = locally compact, Hausdorff. One may associate to X the set VectC(X) of isomorphism classes of complex vector bundles over X. Two vector bundles can be Whitney summed E1 ⊕ E2, making (Vect(X); ⊕) a commutative monoid (the unit being the trivial 0-dimensional bundle over X). Taking the Grothendieck completion of formal differences (which shall be covered later) produces an Abelian group K0(X). In fact, there are higher K-groups: K1(X), K2(X), K3(X), . , making K∗ a cohomology theory. A characteristic feature of K-theory is Bott periodicity: Ki(X) =∼ Ki+2(X) for all i, so in a sense we only have two K-groups to consider: K0(X) and K1(X). As we shall see the construction of the K-groups for a C∗-algebra A follows a similar process (in a way which can be made precise, Swan's Theorem). Vector bundles over X are replaced with projections in the matrix algebras Mm(A). When A = C0(X), the K-theory of the C∗-algebra is naturally isomorphic to the topological K-theory of X. 2. Key Properties: K-theory as a homology theory One may often use K-theory as a `black box', utilising its powerful key properties rather than computing directly from its definition. So we first list some useful such properties, similar to the Eilenberg{Steenrod axioms from topology.1 2.1. Functorality. For each n 2 Z, taking K-theory in degree n defines a covariant func- tor ∗ Kn : C ! A from the category C ∗ of C∗-algebras and the category A of Abelian groups. Being a covariant functor here means the following: ∗ (1) Given a C -algebra A, applying Kn gives an Abelian group Kn(A). 1There is the dual theory to K-theory for C∗-algebras, called K-homology, so I may be using non-standard terminology in this context when listing the properties of K-theory `as a homology theory'. I'm just saying `homology' here because K-theory is a covariant (see below) functor for C∗-algebras. 1 2 INTRODUCTION TO K-THEORY (2) Given a ∗-homomorphism (= homomorphism, in these notes) f : A ! B of C∗-algebras 2 we have an induced map f∗ : Kn(A) ! Kn(B), a homomorphism between the K- groups of A and B. (3) These assignments respect the categorical structures of C ∗ and A . That is: • Identities are respected: if idA : A ! A is an identity morphism then (idA)∗ = idKn(A), the identity morphism on Kn(A). • Composition is respected: if f : A ! B and g : B ! C then (g ◦ f)∗ = g∗ ◦ f∗. 2.2. Homotopy Invariance. We call two homomorphisms f, g : A ! B of C∗ algebras homotopic and write f ∼ g if there is a continuous path ft : [0; 1]×A ! B of homomorphisms with f0 = f and f1 = g (continuity here means that for any a 2 A, the map t 7! ft(a) is continuous from [0; 1] to B). K-theory is homotopy invariant: If f, g : A ! B are homotopic then f∗ = g∗. This implies, for example, that homotopy equivalent C∗-algebras have isomorphic K-theory (we call A and B homotopy equivalent if there exist homomorphisms f : A ! B and g : B ! A with g ◦ f ∼ idA and f ◦ g ∼ idB). Exact sequences. Before stating the next property, recall that a pair of maps f g A −! B −! C is called exact at B if im(f) = ker(g) (we assume that A, B, C, f and g are objects/morphisms in a suitable algebraic category where images and kernels make sense, for us C∗-algebras and Abelian groups). A larger diagram is called exact if it is exact at each position where this makes sense. For example, the following sequence f g 0 ! A −! B −! C ! 0; being exact means that f is injective, g is surjective and im(f) = ker(g). An exact sequence of this particular size is called a short exact sequence. An exact sequence ···! A−1 ! A0 ! A1 ! A2 ! A3 !··· is called a long exact sequence. Up to isomorphism a short exact sequence is of the form q 0 ! I,! B −! B=I ! 0 ∗ where I 6 B, the first map is the inclusion and the second is the quotient. For C -algebras, I here is a (closed, two-sided) ideal of B. In a short exact sequence of groups I is a normal subgroup of B. 2 Being `covariant' means that the induced map goes f∗ : Kn(A) ! Kn(B). The alternative would to be a contravariant functor, which flips the direction of the induced map. For example, topological K-theory is contravariant: if we have a continuous map f : X ! Y between Hausdorff, compact spaces, then we get an induced homomorphism of Abelian groups f ∗ : Kn(Y ) ! Kn(X). This makes sense, the category of commu- tative, unital C∗-algebras is isomorphic to the opposite category of compact, Hausdorff spaces. That K-theory ∗ n is covariant for C -algebras and contravariant for spaces is indicated by the lower (Kn) versus upper (K ) indexing. INTRODUCTION TO K-THEORY 3 2.3. Long exact sequence of an extension. Given a short exact sequence f g 0 ! A −! B −! C ! 0 of C∗-algebras there is an induced long exact sequence g∗ @ f∗ g∗ @ f∗ g∗ @ ··· −! K−1(C) −! K0(A) −! K0(B) −! K0(C) −! K1(A) −! K1(B) −! K1(C) −!· · · : There is some notational laziness here, since we've denoted the maps @ : Kn(C) ! Kn+1(A) identically, without regard for n (of course, we've already done this for the induced maps f∗, g∗ too). By Bott periodicity (a `special' property of K-theory, so we will list it later) the above infinite long exact sequence can be conveniently repackaged as a six-term exact sequence: f∗ g∗ K0(A) K0(B) K0(C) @1 @0 K1(C) K1(B) K1(A) g∗ f∗ This famous six-term sequence is the source of much of the power of K-theory. A short exact sequence of C∗-algebras is, up to isomorphism, the inclusion of an ideal I into B, followed by the quotient map. So the above means that you can say something about the K-theory of one of I, B or B=I whenever you have some information on the K-theory of the others, together with information on how these are connected by the induced/connecting maps (having trivial groups around often saves the day). The map @0 above is called the exponential map, and @1 is called the index map. ∗ 2.4. Additivity. Given a set fAi j i 2 Ig of C -algebras, there is a natural isomorphism ! M ∼ M Kn Ai = Kn (Ai) : i2I i2I That is: \the K-theory of a direct sum of C∗-algebras is the direct sum of their K-theories". Later, we shall see that K-theory has a related continuity property. 2.5. Excision. I won't cover this one because I don't have the time to cover the relative ∗ K-groups (which take a pair (A; A=I) for a C algebra A with ideal I 6 A). It's usual to introduce the relative K-groups in defining the K-theory of non-unital C∗-algebras. Excision theorems allow one to replace relative K-groups with the usual, non-relative ones. However, it is sometimes more natural to express certain objects as elements of relative K-groups; again, we'll pass by this in this introduction. 3. Other important properties The above properties are what make K-theory a `homology theory' for C∗-algebras. But it has some other important properties too, some of which we've already mentioned: ∼ 3.1. Bott Periodicity. For all n 2 Z we have a natural isomorphism Kn(A) = Kn+2(A). Being `natural' means that the isomorphisms respect the extra structure of induced maps and connecting maps of the long exact sequence. In particular, it allows us to derive the six-term exact sequence from the long exact sequence. 4 INTRODUCTION TO K-THEORY 3.2. Suspension. This really belongs in the last section, as a property of a homology theory, but I wanted to mention Bott periodicity first. We have natural isomorphisms ∼ Kn+1(A) = Kn(SA) where SA is the suspension of A. Given a C∗-algebra A, one may define its suspension as the C∗-algebra (3.1) SA := fφ 2 C([0; 1];A) j φ(0) = φ(1) = 0g: Alternatively, you can of course think instead of functions φ 2 C(S1;A) sending some fixed 3 basepoint of the circle to 0. Or you could think of continuous functions φ: R ! A vanishing at infinity. Suspension is a covariant functor: given a homomorphism f : A ! B of C∗-algebras, we have a homomorphism Sf : SA ! SB between their suspensions. It's obvious how to do this: given φ 2 (C[0; 1];A) just compose with f, so Sf(φ) := f ◦ φ 2 SB.